Degirmenci, MursideYuce, Yilmaz KemalIsler, Yalcin2026-01-242026-01-2420241300-18841304-4915https://doi.org/10.17341/gazimmfd.1241334https://search.trdizin.gov.tr/tr/yayin/detay/1257919https://hdl.handle.net/20.500.12868/5257Purpose: The accurate decoding of individual finger movements is pivotal for advanced prosthetic control. In this study, it is aimed to show the effectiveness statistically significant time-domain features based on ANOVA in the classification of EEG signals of finger movements. Theory and Methods: We tried to differentiate the finger movements EEG signals from the dataset which is an open-available and large motor imaginary EEG dataset. The time domain features were calculated from 21 channels of EEG signals of 8 healthy subjects. The effectiveness of two different feature selection methods which are statistically significance-based feature selection (ANOVA) and Principal Component Analysis (PCA) were investigated to discriminate five finger movements and no motor imagery task condition (NoMT). The feature vectors which including all time domain features, ANOVA-selected time domain features, PCA-selected time domain features, and both ANOVA- and PCA-selected time-domain features were applied to 8 different classifiers. The proposed approaches were investigated for both subject-dependent and subject-independent conditions. Results: The highest classification testing accuracy of 35.8% obtained using ANOVA-selected time-domain features with SVM classifier in subject-independent analysis; whereas, in subject-dependent analysis for 8 subjects, the highest classification testing accuracies achieved between 33.3%-57.5% using ANOVA-selected time-domain features for SVM classifier respectively. Conclusion: The results show that the ANOVA-based feature selection method improves the classification performances. The performances of the subject-dependent classifications are higher than the performance of the subject-independent classifications. Therefore, the subject-dependent results of the proposed model have a big potential, which may pave the way for the design of advanced personalized hand prostheses.trinfo:eu-repo/semantics/openAccessFinger movement classificationElectroencephalogram signals]Machine learningStatistically significanceArticle10.17341/gazimmfd.12413343932-s2.0-85185220699Q21257919WOS:001363134900006Q3